crystal Version 2.10 is out and it brings along various enhancements to many of our features. Working for improving the user experience also includes bug fixing, so check out what is new on crystal.
Please find the updates below split by user type:
crystal users should not worry if they can’t remember exactly how to write a filter value in the conversation: thanks to this recent improvement, the search is going to work even if the words inserted don’t exactly match the filter value.
For example, if we have a series of filter values that goes from "Charles de Gaulle Terminal 1" to "Charles de Gaulle Terminal 5", when a user asks for "Charles de Gaulle terminals", crystal should return all of the filter values, from “Charles de Gaulle 1” to “Charles de Gaulle 5”.
To guarantee a better performance, in fact, the system now does not keep just the results which are closer to the original word, but it also includes the ones in a certain tolerance interval.
In this way it is more probable to retrieve a filter value, even in presence of differences from the original value or missing words.
With this version we worked on significant improvements to the automated forecasting feature. Thanks to the v 2.10 enhancement, we are now able to test the automated forecasting and make decisions based on 150 time series with different properties and lengths: until now we were limited only to 30 time series. This leads to a better accuracy and better stability of the algorithm used.
Other enhancements concern a better forecasting accuracy, thanks to:
- Improvements of the forecast interval estimation.
- Use of a more meaningful relative metric for ranking the forecasting approaches.
- Improvement of several forecasting approaches.
- Improvement of the execution.
We also fixed a bug about the forecast result push notifications that could sometimes not be sent out on Android and iOS.
With v2.10 enhancements, the automated forecasting feature guarantees better performance and reliability.
We fixed a bug that would involve inconsistent notification delivery. This issue is no longer happening.
Some enhancements to crystal iOS were obtained through bug fixing, and in particular:
- 1.A crash on the iOS app happening after handling a notification inside the notification drawer when there was only one notification available is not happening anymore.
- 2.The app is not crashing anymore when the data widgets contain empty values or labels.
- 3.Labels for visualizations (e.g. the values on the x axis of a column chart) are not overlapping anymore when showing long values.
Thanks to some bug fixing activity the Android app is working more efficiently.
- 1.A bug sometimes caused crystal to not read answers through the Text to Speech feature, when the user asked something using the Speech to Text option. Now the problem is solved and crystal should not skip any message.
- 2.Until now, the Android app could crash when trying to scroll up to the conversation history. This is not happening anymore.
As of now, it could happen to see the app crashing while scrolling through the topics on the Dashboard. This issue is now solved and no more crashing is expected.
Sometimes the filter values in the Dashboard would show an alias value instead of the original value. Now this issue is fixed and the original value will be shown.
Thanks to bug fixing, the issue that sometimes would cause an improper export of scaled-up visualization values in PDF and Excel files is now solved.
Due to a bug in crystal app for desktop, the topic names in "gestione avvisi", the Italian version of alerts management, were cut and there was no tooltip to show the whole name. Now the issue is solved and the label is not cut anymore, but is wrapped to the next line and perfectly readable.
During the configuration of a topic, in the filter step, when the user is selecting the default value for a mandatory filter from the dropdown menu, the values will now be shown in alphabetical order. This will improve the user experience and make it easier for the user to find the desired value.
Users will now see topic titles in the topic recap view, on the topic card and in the data source management and entity management drawers, like it is shown in the following image.
With the current release, the performances of some features and interactions were generally improved through code refactoring, and in particular we can rely on faster data retrieval, thanks to performance improvements in the data source components.
Some minor bugs concerning the connection and the use of the Identity Provider Google Workspace for the importing of contacts in the User area are now fixed.
As of now, when configuring a topic with a composition objective, a time filter could be found in the suggested questions, even if it was not specified during the configuration process.
For example, a suggested question to be found could be Show me total sales over the last 3 months, even if this filter was not specified during configuration. This bug is now solved and the presence of time filters during the generation of sample questions is taken into consideration only if they are included in the topic configuration.
Until now, it could happen that, even though the default time range was set for a topic in the configuration step, it was not shown when the topic was retrieved in the Dashboard. With the current release this is not happening anymore.